Modeling Driving States Based on Behavior Spectrum Construction Considering Tunnel Environmental Factors

Modeling and analyzing tunnel driving behavior provides insights into driving behavior characteristics and state identification. Previous studies have primarily extracted single or multiple driving behavior features, neglecting their overall time-varied patterns. This study aimed to develop a driving behavior spectrum that considers the coupling effect of driving behavior time series patterns, drivers’ physiological characteristics, and multidimensional environment factors encompassing acoustic, lighting, traffic volume, and road segment type, and to establish a driving state identification model in tunnels. First, a real vehicle test was conducted to collect data on driving behavior, drivers’ physiology, and tunnel environment, from which 13 variables were extracted. A fuzzy comprehensive evaluation method was then applied to assess the complexity of the tunnel environment. Second, the driving behavior spectrum was created for each driver by introducing a single feature recurrence matrix spectrum radius (SRMSR). Then, the hidden Markov model and the criteria importance through intercriteria correlation weighting method were employed to evaluate and classify the driving states. Finally, the composite feature recurrence matrix spectrum radius (CRMSR) based on SRMSR was derived using the Hadamard product and employed as an input variable for a Light Gradient Boosting Machine driving state identification model. The results indicated that the proposed CRMSR was effective in identifying tunnel driving states, enhancing model accuracy as an input. In addition, the proposed method can pinpoint the critical tunnel zones requiring enhanced safety design based on the identification of driving states. It can be used to monitor and identify risky driving states, providing a data foundation for early warning systems and aiding in tunnel design to enhance overall safety.